[Intel-gfx] [PATCH 2/2] drm/i915: cleanup TLB invalidation code

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Fri Dec 9 12:35:25 UTC 2022


On 09/12/2022 12:04, Tvrtko Ursulin wrote:
> 
> On 09/12/2022 11:33, Andrzej Hajda wrote:
>>
>>
>> On 09.12.2022 11:16, Tvrtko Ursulin wrote:
>>>
>>> On 07/12/2022 17:36, Andrzej Hajda wrote:
>>>> Whole register/bit selection logic has been moved to separate helper.
>>>
>>> Why is missing.
>>
>> ...to clean up mmio_invalidate_full function.
>>
>> Will add.
>>
>>>
>>>> Signed-off-by: Andrzej Hajda <andrzej.hajda at intel.com>
>>>> ---
>>>>   drivers/gpu/drm/i915/gt/intel_gt.c | 136 
>>>> +++++++++++------------------
>>>>   1 file changed, 51 insertions(+), 85 deletions(-)
>>>
>>> Diffstat suggests because more streamlined code. Any other reason?
>>>
>>>> diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c 
>>>> b/drivers/gpu/drm/i915/gt/intel_gt.c
>>>> index f0224b607aa4a7..05520ec3264db8 100644
>>>> --- a/drivers/gpu/drm/i915/gt/intel_gt.c
>>>> +++ b/drivers/gpu/drm/i915/gt/intel_gt.c
>>>> @@ -1003,32 +1003,59 @@ void intel_gt_info_print(const struct 
>>>> intel_gt_info *info,
>>>>       intel_sseu_dump(&info->sseu, p);
>>>>   }
>>>>   -struct reg_and_bit {
>>>> +struct reg_and_bits {
>>>>       union {
>>>>           i915_reg_t reg;
>>>>           i915_mcr_reg_t mcr_reg;
>>>>       };
>>>> -    u32 bit;
>>>> +    u32 bits;
>>>>   };
>>>>   -static struct reg_and_bit
>>>> -get_reg_and_bit(const struct intel_engine_cs *engine, const bool gen8,
>>>> -        const i915_reg_t *regs, const unsigned int num)
>>>> +static struct reg_and_bits
>>>> +get_tlb_inv_reg_and_bits(const struct intel_engine_cs *engine, bool 
>>>> write)
>>>>   {
>>>> +    static const i915_reg_t gen8_regs[MAX_ENGINE_CLASS + 1] = {
>>>> +        [RENDER_CLASS]            = GEN8_RTCR,
>>>> +        [VIDEO_DECODE_CLASS]        = GEN8_M1TCR,
>>>> +        [VIDEO_ENHANCEMENT_CLASS]    = GEN8_VTCR,
>>>> +        [COPY_ENGINE_CLASS]        = GEN8_BTCR,
>>>> +    };
>>>> +    static const i915_reg_t gen12_regs[MAX_ENGINE_CLASS + 1] = {
>>>> +        [RENDER_CLASS]            = GEN12_GFX_TLB_INV_CR,
>>>> +        [VIDEO_DECODE_CLASS]        = GEN12_VD_TLB_INV_CR,
>>>> +        [VIDEO_ENHANCEMENT_CLASS]    = GEN12_VE_TLB_INV_CR,
>>>> +        [COPY_ENGINE_CLASS]        = GEN12_BLT_TLB_INV_CR,
>>>> +        [COMPUTE_CLASS]            = GEN12_COMPCTX_TLB_INV_CR,
>>>> +    };
>>>> +    static const i915_mcr_reg_t xehp_regs[MAX_ENGINE_CLASS + 1] = {
>>>> +        [RENDER_CLASS]            = XEHP_GFX_TLB_INV_CR,
>>>> +        [VIDEO_DECODE_CLASS]        = XEHP_VD_TLB_INV_CR,
>>>> +        [VIDEO_ENHANCEMENT_CLASS]    = XEHP_VE_TLB_INV_CR,
>>>> +        [COPY_ENGINE_CLASS]        = XEHP_BLT_TLB_INV_CR,
>>>> +        [COMPUTE_CLASS]            = XEHP_COMPCTX_TLB_INV_CR,
>>>> +    };
>>>>       const unsigned int class = engine->class;
>>>> -    struct reg_and_bit rb = { };
>>>> +    struct reg_and_bits rb = { .bits = BIT(engine->instance) };
>>>>   -    if (drm_WARN_ON_ONCE(&engine->i915->drm,
>>>> -                 class >= num || !regs[class].reg))
>>>> +    if (GRAPHICS_VER_FULL(engine->i915) >= IP_VER(12, 50))
>>>> +        rb.mcr_reg = xehp_regs[class];
>>>> +    else if (GRAPHICS_VER(engine->i915) >= 12)
>>>> +        rb.reg = gen12_regs[class];
>>>> +    else if (GRAPHICS_VER(engine->i915) >= 8)
>>>> +        rb.reg = gen8_regs[class];
>>>> +
>>>> +    if (drm_WARN_ON_ONCE(&engine->i915->drm, 
>>>> !i915_mmio_reg_offset(rb.reg)))
>>>
>>> I'd prefer user readable message was kept but not a blocker.
>>
>> Tried to avoid changes in refactoring, will change.
>>
>>>
>>>>           return rb;
>>>>   -    rb.reg = regs[class];
>>>> -    if (gen8 && class == VIDEO_DECODE_CLASS)
>>>> -        rb.reg.reg += 4 * engine->instance; /* GEN8_M2TCR */
>>>> -    else
>>>> -        rb.bit = engine->instance;
>>>> +    if (GRAPHICS_VER(engine->i915) < 12 && class == 
>>>> VIDEO_DECODE_CLASS) {
>>>> +        rb.bits = 1;
>>>> +        rb.reg.reg += 4 * engine->instance;
>>>
>>> No reason to drop the comment IMO. It explains things somewhat, or at 
>>> least provides a hint.
>>
>> OK
>>
>>>
>>>> +    }
>>>>   -    rb.bit = BIT(rb.bit);
>>>> +    if (write && GRAPHICS_VER(engine->i915) >= 12 &&
>>>> +        (class == VIDEO_DECODE_CLASS || class == 
>>>> VIDEO_ENHANCEMENT_CLASS ||
>>>> +         class == COMPUTE_CLASS))
>>>> +        rb.bits = _MASKED_BIT_ENABLE(rb.bits);
>>>
>>> This could be else if to not have < 12 followed by explicit >= 12, 
>>> but perhaps it is clearer like this, to signify it's two completely 
>>> separate quirks.
>>>
>>> Also, I would perhaps consider having a local i915 since there's a 
>>> good number of engine->i915, but it's up to you what looks nicer.
>>
>> OK
>>
>>>
>>>>         return rb;
>>>>   }
>>>> @@ -1046,14 +1073,14 @@ get_reg_and_bit(const struct intel_engine_cs 
>>>> *engine, const bool gen8,
>>>>    * but are now considered MCR registers.  Since they exist within 
>>>> a GAM range,
>>>>    * the primary instance of the register rolls up the status from 
>>>> each unit.
>>>>    */
>>>> -static int wait_for_invalidate(struct intel_gt *gt, struct 
>>>> reg_and_bit rb)
>>>> +static int wait_for_invalidate(struct intel_gt *gt, struct 
>>>> reg_and_bits rb)
>>>>   {
>>>>       if (GRAPHICS_VER_FULL(gt->i915) >= IP_VER(12, 50))
>>>> -        return intel_gt_mcr_wait_for_reg(gt, rb.mcr_reg, rb.bit, 0,
>>>> +        return intel_gt_mcr_wait_for_reg(gt, rb.mcr_reg, rb.bits, 0,
>>>>                            TLB_INVAL_TIMEOUT_US,
>>>>                            TLB_INVAL_TIMEOUT_MS);
>>>>       else
>>>> -        return __intel_wait_for_register_fw(gt->uncore, rb.reg, 
>>>> rb.bit, 0,
>>>> +        return __intel_wait_for_register_fw(gt->uncore, rb.reg, 
>>>> rb.bits, 0,
>>>>                               TLB_INVAL_TIMEOUT_US,
>>>>                               TLB_INVAL_TIMEOUT_MS,
>>>>                               NULL);
>>>> @@ -1061,50 +1088,14 @@ static int wait_for_invalidate(struct 
>>>> intel_gt *gt, struct reg_and_bit rb)
>>>>     static void mmio_invalidate_full(struct intel_gt *gt)
>>>>   {
>>>> -    static const i915_reg_t gen8_regs[] = {
>>>> -        [RENDER_CLASS]            = GEN8_RTCR,
>>>> -        [VIDEO_DECODE_CLASS]        = GEN8_M1TCR, /* , GEN8_M2TCR */
>>>> -        [VIDEO_ENHANCEMENT_CLASS]    = GEN8_VTCR,
>>>> -        [COPY_ENGINE_CLASS]        = GEN8_BTCR,
>>>> -    };
>>>> -    static const i915_reg_t gen12_regs[] = {
>>>> -        [RENDER_CLASS]            = GEN12_GFX_TLB_INV_CR,
>>>> -        [VIDEO_DECODE_CLASS]        = GEN12_VD_TLB_INV_CR,
>>>> -        [VIDEO_ENHANCEMENT_CLASS]    = GEN12_VE_TLB_INV_CR,
>>>> -        [COPY_ENGINE_CLASS]        = GEN12_BLT_TLB_INV_CR,
>>>> -        [COMPUTE_CLASS]            = GEN12_COMPCTX_TLB_INV_CR,
>>>> -    };
>>>> -    static const i915_mcr_reg_t xehp_regs[] = {
>>>> -        [RENDER_CLASS]            = XEHP_GFX_TLB_INV_CR,
>>>> -        [VIDEO_DECODE_CLASS]        = XEHP_VD_TLB_INV_CR,
>>>> -        [VIDEO_ENHANCEMENT_CLASS]    = XEHP_VE_TLB_INV_CR,
>>>> -        [COPY_ENGINE_CLASS]        = XEHP_BLT_TLB_INV_CR,
>>>> -        [COMPUTE_CLASS]            = XEHP_COMPCTX_TLB_INV_CR,
>>>> -    };
>>>>       struct drm_i915_private *i915 = gt->i915;
>>>>       struct intel_uncore *uncore = gt->uncore;
>>>>       struct intel_engine_cs *engine;
>>>>       intel_engine_mask_t awake, tmp;
>>>>       enum intel_engine_id id;
>>>> -    const i915_reg_t *regs;
>>>> -    unsigned int num = 0;
>>>>       unsigned long flags;
>>>>   -    if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50)) {
>>>> -        regs = NULL;
>>>> -        num = ARRAY_SIZE(xehp_regs);
>>>> -    } else if (GRAPHICS_VER(i915) == 12) {
>>>> -        regs = gen12_regs;
>>>> -        num = ARRAY_SIZE(gen12_regs);
>>>> -    } else if (GRAPHICS_VER(i915) >= 8 && GRAPHICS_VER(i915) <= 11) {
>>>> -        regs = gen8_regs;
>>>> -        num = ARRAY_SIZE(gen8_regs);
>>>> -    } else if (GRAPHICS_VER(i915) < 8) {
>>>> -        return;
>>>> -    }
>>>> -
>>>> -    if (drm_WARN_ONCE(&i915->drm, !num,
>>>> -              "Platform does not implement TLB invalidation!"))
>>>> +    if (GRAPHICS_VER(i915) < 8)
>>>>           return;
>>>>         intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL);
>>>> @@ -1114,33 +1105,15 @@ static void mmio_invalidate_full(struct 
>>>> intel_gt *gt)
>>>>         awake = 0;
>>>>       for_each_engine(engine, gt, id) {
>>>> -        struct reg_and_bit rb;
>>>> +        struct reg_and_bits rb = get_tlb_inv_reg_and_bits(engine, 
>>>> true);
>>>
>>> Ugh so actually what was a once per invalidation lookup is now 
>>> repeated per engine, times two. I wonder if we can do this better. 
>>> Lets think about it a bit.
>>
>> It was always twice, see below.
>>
>>>
>>> Regards,
>>>
>>> Tvrtko
>>>
>>>>             if (!intel_engine_pm_is_awake(engine))
>>>>               continue;
>>>>   -        if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50)) {
>>>> -            u32 val = BIT(engine->instance);
>>>> -
>>>> -            if (engine->class == VIDEO_DECODE_CLASS ||
>>>> -                engine->class == VIDEO_ENHANCEMENT_CLASS ||
>>>> -                engine->class == COMPUTE_CLASS)
>>>> -                val = _MASKED_BIT_ENABLE(val);
>>>> -            intel_gt_mcr_multicast_write_fw(gt,
>>>> -                            xehp_regs[engine->class],
>>>> -                            val);
>>>> -        } else {
>>>> -            rb = get_reg_and_bit(engine, regs == gen8_regs, regs, 
>>>> num);
>>
>> Here is the 2nd call, from old code.
>> Since there are two separate loops there are two calls, caching call 
>> results would be overkill IMO.
>> Or I can put back whole logic to mmio_invalidate_full, GEN12 quirk is 
>> needed only in 1st loop (write), the only redundancy will be with GEN8 
>> quirk, which could be handled with some helper.
>> Is it worth trying? I guess it is no big gain.
> 
> Yes it was always twice in get_reg_and_bit but not the whole register 
> table selection.
> 
> We have some checkes which are per platform, and some which are platform 
> and engine. I propose to keep them split. I made a stab at it like this:
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c 
> b/drivers/gpu/drm/i915/gt/intel_gt.c
> index 0377e1b25be9..d907b9005dd6 100644
> --- a/drivers/gpu/drm/i915/gt/intel_gt.c
> +++ b/drivers/gpu/drm/i915/gt/intel_gt.c
> @@ -988,33 +988,50 @@ void intel_gt_info_print(const struct 
> intel_gt_info *info,
>          intel_sseu_dump(&info->sseu, p);
>   }
> 
> -struct reg_and_bit {
> +struct inv_reg {
>          union {
>                  i915_reg_t reg;
>                  i915_mcr_reg_t mcr_reg;
>          };
> +};
> +
> +struct reg_and_bit {
> +       struct inv_reg reg;
>          u32 bit;
>   };
> 
>   static struct reg_and_bit
> -get_reg_and_bit(const struct intel_engine_cs *engine, const bool gen8,
> -               const i915_reg_t *regs, const unsigned int num)
> +get_reg_and_bit(const struct intel_engine_cs *engine,
> +               const i915_reg_t *regs, const unsigned int num,
> +               bool write)
>   {
> +       struct drm_i915_private *i915 = engine->i915;
>          const unsigned int class = engine->class;
>          struct reg_and_bit rb = { };
> 
> +       BUILD_BUG_ON(sizeof(rb.reg.reg) != sizeof(rb.reg.mcr_reg));
> +       BUILD_BUG_ON(!__builtin_types_compatible_p(typeof(rb.reg.reg.reg),
> +                                                  
> typeof(rb.reg.mcr_reg.reg)));
> +
>          if (drm_WARN_ON_ONCE(&engine->i915->drm,
>                               class >= num || !regs[class].reg))
>                  return rb;
> 
> -       rb.reg = regs[class];
> -       if (gen8 && class == VIDEO_DECODE_CLASS)
> -               rb.reg.reg += 4 * engine->instance; /* GEN8_M2TCR */
> +       rb.reg.reg = regs[class];
> +
> +       if (GRAPHICS_VER(i915) < 12 && class == VIDEO_DECODE_CLASS)
> +               rb.reg.reg.reg += 4 * engine->instance; /* GEN8_M2TCR */
>          else
>                  rb.bit = engine->instance;
> 
>          rb.bit = BIT(rb.bit);
> 
> +       if (write && GRAPHICS_VER(i915) >= 12 &&
> +           (engine->class == VIDEO_DECODE_CLASS ||
> +            engine->class == VIDEO_ENHANCEMENT_CLASS ||
> +            engine->class == COMPUTE_CLASS))
> +               rb.bit = _MASKED_BIT_ENABLE(rb.bit);
> +
>          return rb;
>   }
> 
> @@ -1031,14 +1048,16 @@ get_reg_and_bit(const struct intel_engine_cs 
> *engine, const bool gen8,
>    * but are now considered MCR registers.  Since they exist within a 
> GAM range,
>    * the primary instance of the register rolls up the status from each 
> unit.
>    */
> -static int wait_for_invalidate(struct intel_gt *gt, struct reg_and_bit rb)
> +static int
> +wait_for_invalidate(struct intel_gt *gt, struct reg_and_bit rb, bool mcr)
>   {
> -       if (GRAPHICS_VER_FULL(gt->i915) >= IP_VER(12, 50))
> -               return intel_gt_mcr_wait_for_reg(gt, rb.mcr_reg, rb.bit, 0,
> +       if (mcr)
> +               return intel_gt_mcr_wait_for_reg(gt, rb.reg.mcr_reg, 
> rb.bit, 0,
>                                                   TLB_INVAL_TIMEOUT_US,
>                                                   TLB_INVAL_TIMEOUT_MS);
>          else
> -               return __intel_wait_for_register_fw(gt->uncore, rb.reg, 
> rb.bit, 0,
> +               return __intel_wait_for_register_fw(gt->uncore,
> +                                                   rb.reg.reg, rb.bit, 0,
>                                                      TLB_INVAL_TIMEOUT_US,
>                                                      TLB_INVAL_TIMEOUT_MS,
>                                                      NULL);
> @@ -1068,6 +1087,7 @@ static void mmio_invalidate_full(struct intel_gt *gt)
>          };
>          struct drm_i915_private *i915 = gt->i915;
>          struct intel_uncore *uncore = gt->uncore;
> +       const bool mcr = GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50);
>          struct intel_engine_cs *engine;
>          intel_engine_mask_t awake, tmp;
>          enum intel_engine_id id;
> @@ -1076,7 +1096,7 @@ static void mmio_invalidate_full(struct intel_gt *gt)
>          unsigned long flags;
> 
>          if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50)) {
> -               regs = NULL;
> +               regs = (i915_reg_t *)xehp_regs;
>                  num = ARRAY_SIZE(xehp_regs);
>          } else if (GRAPHICS_VER(i915) == 12) {
>                  regs = gen12_regs;
> @@ -1104,28 +1124,15 @@ static void mmio_invalidate_full(struct intel_gt 
> *gt)
>                  if (!intel_engine_pm_is_awake(engine))
>                          continue;
> 
> -               if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50)) {
> -                       u32 val = BIT(engine->instance);
> -
> -                       if (engine->class == VIDEO_DECODE_CLASS ||
> -                           engine->class == VIDEO_ENHANCEMENT_CLASS ||
> -                           engine->class == COMPUTE_CLASS)
> -                               val = _MASKED_BIT_ENABLE(val);
> -                       intel_gt_mcr_multicast_write_fw(gt,
> -                                                       
> xehp_regs[engine->class],
> -                                                       val);
> -               } else {
> -                       rb = get_reg_and_bit(engine, regs == gen8_regs, 
> regs, num);
> -                       if (!i915_mmio_reg_offset(rb.reg))
> -                               continue;
> -
> -                       if (GRAPHICS_VER(i915) == 12 && (engine->class 
> == VIDEO_DECODE_CLASS ||
> -                           engine->class == VIDEO_ENHANCEMENT_CLASS ||
> -                           engine->class == COMPUTE_CLASS))
> -                               rb.bit = _MASKED_BIT_ENABLE(rb.bit);
> -
> -                       intel_uncore_write_fw(uncore, rb.reg, rb.bit);
> -               }
> +               rb = get_reg_and_bit(engine, regs, num, true);
> +               if (!i915_mmio_reg_offset(rb.reg.reg))
> +                       continue;
> +
> +               if (mcr)
> +                       intel_gt_mcr_multicast_write_fw(gt, rb.reg.mcr_reg,
> +                                                       rb.bit);
> +               else
> +                       intel_uncore_write_fw(uncore, rb.reg.reg, rb.bit);
>                  awake |= engine->mask;
>          }
> 
> @@ -1144,16 +1151,10 @@ static void mmio_invalidate_full(struct intel_gt 
> *gt)
>          intel_gt_mcr_unlock(gt, flags);
> 
>          for_each_engine_masked(engine, gt, awake, tmp) {
> -               struct reg_and_bit rb;
> -
> -               if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50)) {
> -                       rb.mcr_reg = xehp_regs[engine->class];
> -                       rb.bit = BIT(engine->instance);
> -               } else {
> -                       rb = get_reg_and_bit(engine, regs == gen8_regs, 
> regs, num);
> -               }
> +               struct reg_and_bit rb =
> +                       get_reg_and_bit(engine, regs, num, false);
> 
> -               if (wait_for_invalidate(gt, rb))
> +               if (wait_for_invalidate(gt, rb, mcr))
>                          drm_err_ratelimited(&gt->i915->drm,
>                                              "%s TLB invalidation did 
> not complete in %ums!\n",
>                                              engine->name, 
> TLB_INVAL_TIMEOUT_MS);
> 
> So only questions which vary per engine are asked in the engine loops.
> 
> A bit hacky with asserts i915_reg_t and i915_mcr_reg_t are the same 
> underlying type really but may be passable. See what you think.

Or even store register and values (write/read) in struct intel_engine_cs 
at engine init time?

Regards,

Tvrtko


More information about the Intel-gfx mailing list